i7 920 - Core Temperature Differences

S

sn0man

Gawd
Joined
Jan 25, 2007
Messages
609
Just finished an i7 build and I'm a little curious about temperature differences across the cores.


i7 920 @ stock / Vigor Monsoon III / Arctic Silver 5
Asus P6T (Bios settings Auto)
OCZ 1600 3x 2GB @ stock
Asus HD 4830 512MB @ stock
Antec P182 Case
Corsair CMPSU-850TX Power


When first turning on the machine, temperatures are as follows:

2mxqxk7.jpg


After about an hour of Intel Burn Test:

2mqw7sh.jpg


21edpoi.jpg




From what I understand, the Antec P182 isn't a monster in terms of cooling. In addition, the 4830 seems to be running around 50C and doesn't exhaust from the back of the case.

Given this setup do the temperatures seem right?



Also, for temperatures:
Cores 1 & 2 seem to run pretty close to each other. Core 0 runs a few degrees hotter, while core 3 runs several degrees cooler than everything else.

Does this look normal or seem a bit off?


I used the credit card / thin spread method for the Arctic Silver rather than dot in the middle or line.

Given everything above, what are your opinions on whether I should clean and reseat the cooler using another method?
 
Nah.. looks pretty normal to me... my core0 runs about 2 to 4c higher than the rest and cores1 - 3 can vary by 2 or 3 also.. did you use AS5 or something which require a "set" period? if so, let the thermal compound set for a couple days at least before getting too nervous.. if it gets worse maybe reseat it.. but I don't think it looks too out of the ordinary.
 
the varying temps is a known issue with intel processors.. usually bad sensors.. just stick with the highest core temp you see and use that as a reference for the rest of them..

and stick with real temp for your cpu temps.. since its setup for the i7.. the rest of them arent updated yet to support the i7 so the temps will look way off on all the other programs compared to real temp..
 
On my i7-920 cores 1 and 2 run 5-6 degrees hotter then 3 and 4. When I stress test all cores top out within 1-2 degrees of each other. This was the same with the stock intel cooler and tim and when I replaced it with a CM V8 with MX-2.

Tark
 
sn0man said:
Given this setup do the temperatures seem right?
Click to expand...
Considering you're running at stock speeds, it isn't out of the ordinary.
sn0man said:
Also, for temperatures:
Cores 1 & 2 seem to run pretty close to each other. Core 0 runs a few degrees hotter, while core 3 runs several degrees cooler than everything else.

Does this look normal or seem a bit off?
Click to expand...
Perfectly normal.
sn0man said:
I used the credit card / thin spread method for the Arctic Silver rather than dot in the middle or line.

Given everything above, what are your opinions on whether I should clean and reseat the cooler using another method?
Click to expand...
Might as well just leave it unless you start encountering temperature issues in the future.
 
My curiosity got the best of me, and not to leave things be, I decided I'd try lapping and reseating anyway.

Final products of my lapped CPU / CPU Cooler:

2nm2hoj.jpg


11log2e.jpg


rclwnq.jpg



I tried only using a dot of AS5 in the middle of the CPU this time, rather than spreading it thinly across like last time.

Temperature results are pretty much unchanged - including the spread from core #0 to core #3.

It's a bit disappointing considering the effort to remove everything and time taken to lap / put it all back together. IF there was an improvement from what I did, I'm guessing it's a wash due to ambient case temps (heat from the video card?).

Now that I've got that settled, I guess I'm clear to start overclocking...
 
what are good temps I just built 2.66 i7 with a mugen 2 and I run 40C idle gotten to 60C under intel burn load.
 
the thing you arent understanding is that intels die sensors are a pos and fail under 50C.. so anything under that is completely unreliable as a legit temp.. try overclocking then see what the difference is.. and like i said your core 3 sensors borked.. just disregard the reading on that core.. doesnt mean the cpu's messed up.. just that sensor doesnt work correctly..

also the dot of as5 probably did more harm then good.. id say check to make sure it actually spread across the whole cpu..
 
sirmonkey1985 said:
the thing you arent understanding is that intels die sensors are a pos and fail under 50C.. so anything under that is completely unreliable as a legit temp..
Click to expand...
The sensors that Intel uses in the Core i7 CPUs are greatly improved compared to the Core 2 sensors, and are able to report low temperatures with significantly greater accuracy as well as not having the stuck sensor issues that plague many 45nm Core 2 CPUs.
 
Zero82z said:
The sensors that Intel uses in the Core i7 CPUs are greatly improved compared to the Core 2 sensors, and are able to report low temperatures with significantly greater accuracy as well as not having the stuck sensor issues that plague many 45nm Core 2 CPUs.
Click to expand...


yeah i know.. except this specific core hes using has both of those issues.. :p
 
I know the temps are high. Obviously air flow is going to have to be improved in my setup (and voltages lowered).


After running @ 3.8GHz stable, 1.2875v, I decided to up the voltage and check how the temperatures increased during stress.


The screens are from a 1hr+ run of prime95, while also working online doing some reading, etc. No errors, everything solid, just stupidly hot.

@4.0 GHz, 200 x 20, DDR 1600, uncore 3200 MHz, QPI 3600 MHz, HT [on], Turbo/Speedstep/C1E/C-State/TM [off], Asus LLC [on], skews [off/auto]

CPU: 1.38125 bios / 1.376 idle / 1.392 load
DRAM: 1.66
QPI/DRAM: 1.38125
QPI/PLL: 1.90

2aewdg1.jpg


Idle temperatures with those settings were around 37-44 C.

I was getting into windows @ 4 GHz with much less on the vcore (~1.35) but eventually getting blue screens in prime95. This was also before I started changing the QPI/DRAM and QPI/PLL voltages much though - so I'm hoping I can find a sweet spot for those which allow me to drop the vcore back and still be stable.

If anyone has more reasonable figures for running something similar on a P6T vanilla (or any of the variations), I'd love to hear some suggestions on specifics. The processor itself is a 3835A batch.




On to the case...

14joqq8.jpg


  • Antec P182
  • Bottom fan between hard drives & power supply removed for space (will probably re-route some wiring and get it back in soon).
  • Push/Pull CPU cooler setup, with 1 case exhaust fan in line with it. The 2nd case exhaust fan is at the top of the case, above the CPU cooler.
  • Asus HD 4830, single-slot, no exhaust.

Suggestions on case setup?

People have placed the bottom fan (which I removed) in FRONT of the hard drives with a few mods, but with the fan in the middle of the bottom chamber it's basically a wind tunnel anyway.

No exhaust on the video card sucks, but how much would you imagine that hurts case temperatures when overclocking a CPU that gets as hot as the i7?

Also, the top exhaust fan... should I switch it to intake? With the majority of the fans exhausting air I guess the negative pressure is good, but I honestly feel VERY little air coming out of the top vent. In addition it's nowhere near as warm as the air coming from the case's rear exhaust.



Lot of info and lots more questions, I know. Thanks for anyone who's got opinions and would like to offer them.
 
Well I can tell you I also thought my 3.8 oc was stable at 1.3v and was prime stable for 24 hours. Didn't really feel like it was stressing the CPU as browsing the net and doing daily tasks was not effected. Also ran into a thread about running multiple stress tests. Tried running Linx and Prime95 together and BSOD with in a few minutes. Need to raise the CPU voltage to 1.35v to make it stable. With that I am able to run both and ATI tool, and it really bogs the system down. Took me almost 2 min to get the monitor out of sleep state, so I know all resources were being used. I ran it for a few hours like that with no errors.

Might want to be something you play around with. You don't want silent corruption sneaking up on you.
 
I had mine at 4.02 on Asus P6T @ 1.35v. It ran Linpack and Prime 95 (separately) stable for hours. Asus Probe never went above 75 C. I started playing left 4 dead and it crashed an hour into the game. Upon reboot it crashed about 5 mins into it and I could not run a linpack or prime95 successfully.

I have since backed down to 3.8 @ 1.35 and everything is stable again. Weird thing is my temps seem to have jumped. Now when I run Prime95 I get 79 C...sometimes 80 c. Using a Scythe Mugen 2 cooler.
 
groebuck said:
I had mine at 4.02 on Asus P6T @ 1.35v. It ran Linpack and Prime 95 (separately) stable for hours. Asus Probe never went above 75 C. I started playing left 4 dead and it crashed an hour into the game. Upon reboot it crashed about 5 mins into it and I could not run a linpack or prime95 successfully.

I have since backed down to 3.8 @ 1.35 and everything is stable again. Weird thing is my temps seem to have jumped. Now when I run Prime95 I get 79 C...sometimes 80 c. Using a Scythe Mugen 2 cooler.
Click to expand...

maybe your heatsink got bumped or something? that could break the good set seal the thermal compound had.. just a guess... when your machine crashed did you fiddle with the advanced cpu features when lowering the OC? I am thinking specifically about Hyper Threading, toggling that on when it was off before can give you a 10c shift up in temps across all cores.. just another shot in the dark guess.
 
You must log in or register to reply here.
Back
Top